首页资源分类嵌入式系统 > Thermal Management in PowerPC Microprocessor

Thermal Management in PowerPC Microprocessor

已有 445125个资源

下载专区

文档信息举报收藏

标    签:ThermalManagementinPowerPCMicroprocessor

分    享:

文档简介

Thermal Management in PowerPC Microprocessor

Mu1tichip Modules Applications

文档预览

Thermal Management i n PowerPC Microprocessor Mu1 tichip Modules Applications Tsorng-Dill Yuan I BM Microelectronics Division 1580 Route 52 I.Iopewel1Junction, New York 12533 *l‘licniial n~:uiagcmcnt considcrations of for I’owcrl’C microproccssor multichip modulcs were discusscd in this papcr. I;low bypass effects on both I’owcrFC 605 and I’owerPC 604 MCM heat sink arid modulc thcrinal performance were quantified by computation~flluid dynamics and conjugate conduction analysis. In PowerPC 603 MCM application, thc heat sink thermal resistance can vary as large as 80 % due to such effect. while in PowerPC 604 MCM application, the variation in thermal rcsistance can be as large as 100 %. The conjugate heat transfcr features in both PowerPC 603 and PowerPC 604 MCM were examined. Results show that the percentage of module power entering into the heat sink seems quite independent on the module component thermal properties. In general, about 70 % of the module heat entering into the heat sink for PowerPC 603 MCM and about 85 % of the module heat cntering into the heat sink for PowerPC 604 MCM. Both module thermal resistances and the percentage of chip and module power distributions are affected by those material properties. Their effects werc studied to understand the MCM thermal characteristics and to defme the critical design parameters for thermal perforrnance optimization. Inlrodiiction Thermal management considcrations of multichip modules have been reviewed by Bar-Cohen [1J Chu et a1.[2] and Nakayama [SI with focus on cooling technologies in high pcrformance main frame computer applications. Recently, thermal management challenges are moving toward thermal control in low cost au cooled solutions. Mok [4] gave insights in some technology aspects of a viable cooling solution for MCM. Ixc et al. [SI emphasized on the modclling tcchniquc by showing the usability of a computational fluid dynamics mcthod in temperature predictions. As the electronic trcnds becomes emphasizing on low cost or personal-based computers, the MCM thermal management challenges is now how to accurately predict the temperature of a heat sinked electronic module in a conjugate heat transfer environment. It was noted previously by Lee [SI and Wirtz et al. [7] that the heat sink thermal resistance in a realistic environment is highly dependent o n the air flow bypass over the heat sink. Such effect is governed by the tip clearance and spanwise spacing in thc systcni cross section dimensions. Lee [SI and Wirtz et al. [7] also tried to quantify its effect on the heat sink performance prediction using analytical approdies. By using computational fluid dynamics approach Yuan [ 8 ] quantified the flow bypass effect on hcat sink thermal performance. I Iowcver, such effccts have not been apptied to the prediction of electronic module thermal performance. As the elcctronic packaging moves toward the surface mount technology, the printed circuit card can now help to dissipate the heat. Therefore, only a portion of the heat generated by the module is dissipated through the heat sink. The challenge lies in the way to quantify the amount of the module power which are carried away by the heat sink. Furthermore it is important to understand how the MCM thermal performance is dependent on the various thermal propertics of module substrate, cap, grease and card. I’hcse considerations form the motivation of this study. Prohlcm Statcmcnt In ordcr to understand thesc thermal management considerations, PowerI’C 603 and 604 multichip modules, with their practical applications, were chosen for illustration. The problem of intercst focuses o n both heat sink and MCM module thermal performances. The heat sink per- formance study will be focusing on the flow bypass effects, as an extension of ear!ier study by Yuan [SI. The results will be applied to both PowerPC 603 and 604 MCM to study thcir effects on MCM module thermal pcrformance. In addition, the effects of the various coinponcnt tlicrinal properties 0-7803-3793-X/97/$5.0001997 IEEE 247 Thirteenth IEEE SEMI-THERMTMSymposiutn on MCM t I w i ~ i ~ ~ :picl :rfi)rrnance characteristics will bc st udictl. Ilolli tnodulc thctinnl resistances and thc pcrcciiliigc o l chip and module power distributions iis cllcc~c~bly thosc material properties will bc cx;iiiiiricd, I’owcrl’C 60.3 and 604 microprocessors C9) uscd 32-hit o f I’owcrl’C architccture. PowerPC 603 rnicroproccssor was used for the cost sensitive, dcsktop and poi-tddc pcrsonal computer systems whilc I’o\vcr I’C 604 microprocessor was used for high pcrforiiiniicc uniprocessor or multiprocessor in dcsktop pcrson:il computers, workstations and scrvcrs. 1:igrirc I shows the clup layout both l’owcrl’(: 603 : i d 604 MCMs on ;I 44 mm x 44 intn ceramic coluinn grid array module. The dies arc all flipchip joined to the substrate. A nonlicrmctic aluriiinuni cap is used to encapsulate thc h4CM modulc w i t h scalant.and grease. ‘I‘hc module is populated with a PowerPC chip, a 512 KII of data RAM chip, an associated tag R A M and cachc control chip, a clock distribution chip and a I’CI bridgc chip set. Table 1 shows thc chip s i x arid power information for both MCMs. ‘l’licpowcr of the PowerPC 603 chip is 2.2 watt and 3 watt at the nominal and maximum conditions while the power of the I’owerI’C 604 chip is 10 watt and 13 watt at the nominal and maximum conditions. For PowerPC 603 MCM, the total module power is 7.8 watt and 10.1 watt and for PowerPC 604 MCM, the total power is 15.6 watt and 20.1 watt at the nominal and maximum conditions. lieat sink thermal nnnlysis To study the heat sink thermal performance, consider a heat Silk placed in a computer assembly in a thrce dimensional Cartesian coordinate systems shown in figure 2. The heat sink width is w and - height is 11, while the duct width is W and height is H . The tip clearance is thus (I1 h) and the spanwise spacing is 0.5 (W - w), The governing parameters to be studied are the tip clearance (TC), the spanwise spacing (SS), and the air velocity. Assuming incompressible flow. steady flow, and neglecting the gravity forces, the governing equations for mass conservation is CII The governing equations for momcnturn is [*I And the conscrvation of energy equations is written as C31 Equations [1 - 31 together with appropriate boundary conditions constitute the inathcmatical problem, which arc solved by an appropriate numerical solution scheme. The computational procedure and accuracy was tested and verified with the experimental data and the detaijs can be found in Yuan [S]. A4CM thermal model forniulfltion TO study the PowerPC MCM thermal per- formance, a t h e e dimensional f i t e difference con- duction model was formulated. The boundaries on the heat sink surface, the module side, the card top surface and the card bottom surface are modeled as convective thermal boundary conditions. The interconnection layers between the module and the card are modeled as a single layer of solid using an equivalent thermal conductivity. T o realistically consider heat spreading, card assembly and I/O wifing, a 15 tnm shadow space is assumed around the module in the model. Ai heating effects arc considered in this model. Convective boundary conditions were spccified using the heat sink resistance value for heat sink surfacc boundary and and using flat plate correlations data suggested by BarCohen [I] for other convective boundaries. Par-ameiers de$nition For PowerPC 603 MCM, the heat sink dimen- sion is selected with 38.1 x 38.1 x 10.2 m m in size, having 10 straight fins of uniform space with width, t of 1.6 mm, fin height of 7.1 m n and fin gap width, s of 2.5 mm. The tip clearance (TC) is chosen to be varying from 2 mm, 5 im, 10 mm, 20 mm to 30 mm and the spanwise spacing (SS) is varying from 5 mm, 15 I n n , 25 mm to 40 mm. Tlus gives the (AR), which is dceofrirncesdpoaisidhin*gwva/(luWes *ofI{)a,revaarryaitniog from 0.65 for 1C = 2 mm and SS = 5 m m to 0.08 for TC = 30 mm and SS = 40 mm. For PowerPC 604 MCM, the heat sink dhncnsion is 53.3 x 53.3 x 20.0 nim in sizc n-ith 16 straight fins of uniform spncc with width, t of 1.5 mm, fin height of 17.0 mm and fin gap width, s of 1.9 xnm. The tip clearance (TC) is chosen to be varying from 2 mm, 5 inm, 10 mm, 15 m m to 25 mm and the spanwise spacing (SS) is varying from 5 mm, 10 mm, 15 mm, 25 mm to 35 m. This gives the corresponding values of area ratio (AR), which is defined as h * w /( W * I€),varying from 0.765 for ‘I‘C = 2 mm and SS = 5 mm to 0.19 for T C = 25 mm and SS = 35 mm. ‘The air flow approach velocity, V,,is 1, 2 and 3 m/s. Heat sink performnnce results 248 Thirteenth IEEE SEMI-THERMTMSymposium 'I*lichc:it sink thcrmal performance for both I'owcriY: 603 and 604 MCM were presented. 1;igurc 3 coiiip:ws tficrinal resistances and pressurc drops in I I I ~ctiuironmcnt with 'TC = 2 mm and SS = IS niiii. At I tn/s velocity, hcat sink resistance is S.OX (:/w:ind 1.56 C/w for 603 and 604 hcat sink rcspcctivcly. As shown Iatcr at 1 m/s velocity tlie Iicat sink rcsist;incc is 9.1 C/w and 3.56 C/w for 603 arid 604 licat sink respectivcly in an open environmcnt. 'I'his incatis thcre is about 80 70and 130 % diflcrcncc in thcrinal resistance magnitude due to thc cxislcncc o f flow bypass. 'I'hc ilow bypass effccts on the pressure drop and thcrmal rcsistances for PowerPC 603 and I'owcrlY: 604 hcat sinks for various TC and SS valucs at I m/s air arc shown in figures 4, 5, 6 and 7. In I'owcrI'C 603 MCM hcat sink we can see that as hrgc RS 80 Yo variation in thennal resistance bctwccn 'I'C = 2 nitn and SS = 5 mtn and 1'C = 30 mtn arid SS 40 mm exists. The corre- sponding prcssure drop variation is about 10 timcs bctwccii ' I C = 2 m m and SS = 5 mm and 'TC = 30 m m and SS = 40 mm. In I'owerPC GO4 MCM Iicnt sink, thc variation in thermal resistance bctwccri 'I'(: = 2 tntn and SS = 5 mtn and '1'C = 30 mtn arid SS = 40 m n is about 100 % and tlie corrcsponding prcssure drop variation is also about I O tinics. -1'liesc findings indicate how important it is to bctter control the flow bypass for MCM thennal management. h iC M modi& performnnce discussion The MCM thermal performatice due to the flow bypass effccts are shown in figures 8 and 9. The MCM module thermal performance car1 be defined as the MCM module thermal resistance which is the tcmpcrature difference between the PowerPC chip and the ambient air dividcd by the module power. We cat1 then determine the clup temperature by adding the ambient air temperature with the multiply of the module power and the module thermal resistance, as shown in figures 8 and 9. In PowerPC 603 MCM, with its module nominal power of 7.8 watt, it was found that maximum temperature difference due to the flow bypass effect can be as large as 12 degree C or 36 "/o in module thennal performancc In PowerPC 604 MCM, with its module nominal power of 15.6 watt, the maximum temperature ditrerence due to the flow bypass effect is 15 degrcc C or 42 YOin module thermal performance, PowerPC A/ICM parametric perfoi-niancc discussion T h e DowerI'C MCM thermal performance can be affected by thermal conductivity values of module cap, thcrmal greasc, substrate arid the effcctive thermal conductivity of printed circuit card. Due to heat sprcading effcct, thc module cap thickness is also considcrcd as a paramcter. in this study thermal conductivity of the cap is varying from 5, 21, 100, 180 and 350 W/ m K. Thermal conductivity of the substrate is varying from 0.3, 5, 21 and 140 W / m K. Thermal conductivity of the thermal greasc is varying from 0.5, 1.5, 2.8 and 3.8 W/m K. The effective thermal conductivity of the printed circuit depends o n the number of power and signal layers and its value is varying from 1 , 3, 13 and 20 W,/m K. The cap thickness is vaiying from 1, 2, 3, 4 and 6 mm. The baseline heat sink resistance for PowerPC 603 MCM is 4.2 C/w obtained at air velocity of 1.5 m/s with TC = 2 mm and SS = 15 mm. 'The baseline heat sink resistance for PowerPC 604 MCM is 1.56 C/w obtained at air velocity of 1 m/s with TC = 2 mm and S S = 15 mm. In the MCM model the original values of thermal conductivity for cap, substrate, card and grease are 180 WJ mK, 21 W/ mk, 3 W /mk and 2.8 Wi ~nkr,espectively. The original value of cap tllickriess is 3 mm. 'The rnodule thermal resistances at different substrate and cap thermal conductivities for both I'owerI'C GO3 arid 604 MCM are shown in figure 10. The effect of substrate thermal conductivity is less than that of the cap. By changing substrate K from 21 to 140 W/m k only improve 3 YOthermal performance in I'owerPC 603 MCM and 12 Yo in PowerPC 604 MCM. The cap thermal conductivity plays a sigrlificant role in affecting the thermal performance. For PowerPC 603 MCM, the change in cap K from 5 W/mk to 180 W/mk gives 26 YO thermal improvement, while in PowerPC 604 MCM the change in cap K from 5 W/mk to 180 W/mk gives 100 YOthermal improve- ment. The difference is probably due to the different power rates between these two MCM. Figure 11 shows the module thermal resistances at different thermal conductivities of printed circuit card and thermal grease for both PowerPC 603 and 604 MCM. As card K changes from 1 W/mk to 20 W/mk, it is note that, only 7 Yo thermal improvcment in PowerPC 604 MCM but 17 Yo thcrmal improvement in PowerPC 603 MCM. This shows that, in PowerPC 604 MCM, most of the heat was carricd away by the heat sink and its thermal performance was less affected by the card K. As thcrmal grease K changes from 0.5 W/mk to 2.8 W/mk. thcre is 20 YO thermal improvement in PowerPC 603 MCM and 61 YO thermal improvement in PowerPC 604 MCM. This indicates that PowerPC 604 MCM, with its relatively higher power, depends more on thermal grease conduction to improve its thermal perfonnance. Figure 12 shows the module thermal resistances at different cap thickness. Using aluminum cap with K = 180 W/mk, the cap thickncss is varied from 1 mm to 6 mm. Change in cap thickness from 1 mm to 6 mm gives 9 % thermal 249 Thirteenth IEEE SEMI-THERMTMSymposium itnprovciiicnl in I”wr1’C 60.7 MCM and 24 YO thermal iii\provcinciit in I’owcrI’C 604 MCM. ’l‘his shows ilia1 I’owcrI’C 604 MCM, with its h i g h hcaf Iltlx, nccds bcttcr tlierrnal spreading due to tlw cap Iliickncss t o dissipate heat into the hcat sink. In M C M thcrtnal tnanagcment, as some of the hcat is dissipnfing tlirough thc substratc into the card, it is important to know the amount of the rnodulc hcilt riitc cariictl away by tlie heat sink. ‘To hcttcr uiiticxxtnnd the conductive characteristics, two paramctcrs which are tlie pcrccntage power of thc I’owcri’C: chip cntcring into thc cap and the pcrcciitagc powcr o f thc MCM module powcr cnlcring inlo Ilic h a t sink were used. 1;igure I3 shows thc heat rate distributions at ciikrctit sulxtratc thcmal. conductivities h r both I’owcrI’(: 603 arid 604 hKM. Results show that about 88 ‘‘A of thc module powcr enters into the licat sink for I’owcrI’C 604 MCM and about 71 YO of thc modulc power enters into the hcat sink for I’ov,wI’C 603 M C M at all valucs of substratc K. ‘I‘hc pcrcctitngc of the PowcrPC chip power cntcring into tlie cap depends on the substrate K and its valilc dccrcases sharply as substrate K value increase for !mth MCMs. PowerPC 604 MCM sccnis to hnvc rclatively low percentage of clup powcr cntcring into the cap as substrate K increase, Figure 14 shows the heat rate distributions at different cap thermal conductivities for both I’o\vcrPC 603 and 604 MCM. It is obscrved that as cap K incrcascs, the percentage of PowerPC chip power entcring into the cap increascs strongly, whilc the pcrcentage of module power entering into the lieat sink seems to remain constant. Dramatic increase of the pcrcentage of PowerPC chip power into the cap occurs when the cap K increases from 5 W/mk to 21 W/mk. This fact helps explain why the module thermal resistance will be high at low cap thennal conductivity when only a small portion of the chip power is entering into the cap and heat sink. Figure IS shows the heat rate distributions of both PowerPC 603 and 604 MCM at different card thermal conductivities. Both percentages of PowerPC chip power and modulc power entcring into cap and heat sink decrease as card K increases. As expected, when card K increases, there will be more heat entering into the card, and therefore the amount of the heat entering into the cap and heat sink decreases. As shown in figure 15, the decreasing rate of the percentage for I’owcrPC 603 MCM is stronger than that of I’owcrI’C 604 MCM. T h i s helps explain the reason of the relatively strongcr thcrinal improvcmcnt (17 O h vs 7 YO for PowerI’C 604 M C M ) in PciwcrI’C 603 M C M as printed circuit card K changes from 1 W/mk to 20 W/mk. Figure 16 shows the hcat rate distributions at different grease thermal conductivity for both PowerPC 603 and 604 MCM. As grease K increase, the the percentage of PowerPC clup power entering into the cap increases strongly whiIe thc percentage of module power entering into the heat sink remairis constant. The sharp increase of the percentage of chip power entering into the cap occurs when grease K increase from 0.5 W/mk to 1.8 W/mk. As only small portion of the chip power was entering into the cap due to the IOW grease K value, the module thermal resistance increases. Figure 17 shows the heat rate distributions at different values of cap thickncss for both PoiverI’C 603 and 604 MCM. As the cap thickness increases, the percentage of I’owerPC chip power entering into the cap iticrcases while the percentage of module power entering into the heat sink decreases. The combined effcct help to reduce the module thermal resistance as the cap t h i c k ” increases from 1 mm to 6 inm. As cap thickness contkues increasing from 6 mm, the heat spreading effcct will be reaching its limit and its module thcrmal resistance will start incrcasing due to conduction resistance. In summary, thc percentage of module power into the heat sink can be afkcted by card thermal conductivity and cap thickness. And it is almost unaffectcd by thermal conductivities of cap, substrate and thermal grease. As far as the percentage of chip power entcrhg into the cap is conccrned, its value can be increased by increasing in cap or grease thermal conductivity or cap thickness and its value can be decreased by increasing of card or substrate thermal conductivity. From module ttiermal pcrformance viewpoint, the printed circuit card thermal conductivity has stronger effect on PowcrPC 603 MCM thcrmal pcrformance than the PowerPC 604 MCM. On the other hand, the thermal conductivitics of substrate, cap and grease and the cap thickness have stronger effects on PowerPC 603 MCM thermal perforniance than the PowerPC 603 MCM. Conclruions In electronic module cooling, the external resistance thermal rnanagemcnt must consider flow bypass effects for better thermal performance control. In this study the flow bypass effects on both PowerPC 603 and PowerPC 604 MCM thermal perfonnance are quantiqed. In the case of PowerPC 603 MCM application, the hcat sink thennal resistance can vary as largc as 80 YOdue to such effect. At a fixed module operating condition, the temperature prctliction can be off by 12 degree 250 Thirteenth IEEE SEMI-THERM” Symposium (: simply duc to the flow bypass effect. In tlic case of I’owcxl’C 604 MCM application, the variation ill tlicrmnl rcsistance can be as large as 100 “o/ and Ilic: tcmpcrature prediction can be off by 15 degree (‘ siiiiply duc to the flow bypass effect. Although (i)und ~ n wd rittcn in specific applications, it is felt IIuit such ktiowlcdge is useful for practical thermal tn:~nngcmcntenginccring. ‘I‘hc conjugate heat transfer features in both I’o\rcrl’C 603 and PowerPC 604 MCM were cxnmiticd. Iiesults show that the percentage of rnoriulc powcr entering into the heat sink seems quitc itidcpciidcnt on the module component llicrmnl properties. In general, about 70 YOof the modulc heat entering into the heat sink for I’owcrI’C 503 MCM and about 85 % of the modulc heat cntering into the heat sink for I’owcrI’C 604 MCM. From the modulc thermal pctforrnnncc vicwpoint, it is found that thc cffect of card thcrmal conductivity is stronger in PowerPC 603 MCM while the effect of substrate, cap and thcrmal grease thcnnai conductivity and the cap thickness is stronger in PowerPC 604 MCM. ‘The cffccts of those tnaterial thermal propertics need to hc considcred as the critical paramcters in these hlCM tliennal performance optimization. As external thermal resistance is strongly dcpcndent on the system and environmental conditions, a computational modeling tool need to consider the air flow behavior in order to realistically predict the thermal performance. From tlus consideration, a good computational fluid dynamics model seems to be a necessity rather than an alternative. The effects of the modulc component thermal properties on the MCM thermal pcrformance are shown to be significant. Both external and internal thermal resistances are coupled in a conjugate heat transfer mode. Therefore the tradi- tional concept of internal and external resistances is liinitcd for practical use. Reference 1, I3a.r-Cohen, A., ”Thermal Management of Airand Liquid- cooled Multicllip Modules,” IEEE CIIMT Trans., Vol. CI-IMT-10, No. 2, pp. 159-175, 1987. 2. Chu, R. C. “Heat Transfer in Electronic Systems,” I’roc. 8 th Intl. Heat Transfer Conference, Vol. 1, pp. 293-305, 1986. 3. Nakayama, W., ”Thermal Management of Electronic Equipment: A Review of Technology and Research Topics,” Appl. Mech. Review, Vol. 39, no. 12, pp. 1847-1868, 1986 4. Mok, L. ‘Thermal management of Siliconbased Multichip Modules”, Proceedings - IEEE Semi-Therm Symposium, pp. 59-63, 1994. 5. Lee, T., Chambers, B. and Mahalingam, M., ”Application of CFD Technology to Eiectronic ‘I’hermal Management”, IEEE Trans. Components, Packaging and Manufacturing Technology, Part 13: Advanced Packaging, v. 18, n. 3, pp. 511-520, August, 1995. 6. Lee, S.,”Optimurn Design and Selection of Heat Sinks,“ Eleventh IEEE Semi-therm Symposiums, pp 48-54, 1995. 7. Wirtz, R. A., Chen, W., and Zhou, R., 1994,”Effect of Flow Bypass on the Performance of Lsngitudinal Fin Heat Sinks,“ ASME Journal of Electronic Packaging, Vol. 116, pp. 206-211. 8. Yuan, T.D.“Computational Modelling of Flow Bypass Effects of Heat Sink in a Rectangular Duct”, Semi-Therm, 1996 9. PowerPC 603 RISC Microprocessor User’s Manual, IRM/Motorola, 1994. PoverPC 603 MC!.! PowerPC 604 MCM %idpa Brldge COnbOtiar Buff- DCilaobck. 4Rau Tc9 XAA! Bua Buffer 12.1 9 . 8 ~ 4 . 1 3 . 21~.3 mm mm Chlp powe .r No- Pdoewnaelrv l0/13 watt 1 w6i8mPi 0.6/0.7 0.4,/05 watt wan 1 1 w1?xh7 2 w9..mp62 a 0.7‘0.9 OB/:.O watt 1 4 ?AY -2% 0.5./0.7 02’0.4 watt ! : J;Jg,6v%J Table 1. Chip information for PowcrPC 603 and 604 MCM. 251 Thirteenth IEEE SEMI-THERMTMSymposium PowerPC 603 M C M PowerPC 604 M C M Heat sink resistcnce, C/w 6I 5 4 3 2 1 Pressure drop, Pa 18 0.5 1 1 .s 2 2.5 3 3.5 Air velocity. m/s Poweff'C 603 haotfink 38.1 x 3a1 x 10 mm -c PaworPC 604 haagink 53.3 x 53.3 x 20 mm -6- p ~ ~ a r p60c3 heatrink Rw ......\.ma....d-r w poa&C 60: hw-ink Wusure d w _.....A....... Figure 3. Heat sink resistance and pressure drop at various velocities. 2.8 THERMALGREASE P r e s s u r e dr p . Pascal '3 Figure 1. Power PC 603 and 604 MCM chip layout and module cross section. i -I Air Flow 4 + ' Heat Sink Figure 2. Computational domain of heat sink modeI. Figure 4.Flow bypass effect on PowerPC 603 heat Sink pressure drops. 252 Thirteenth IEEE SEMI-THERMTMSymposium Thermal resistonce, C/w 10 . 6 5 S I 0 5 10 15 20 25 30 35 . Tip clearance. mm - spanwise spacing 5 mm --c spanwise spacing = 15 mm -4- spanwise spacing = 4 0 m m 1 -..-----. two dimensional .-.,.... spanwise spocing = 25 mm ".... Asymptotic free space Figure 5. Plow bypass effect on PowcrPC 603 heat sink tcsistances. ?ressure drop, PA Heat sink thermal resisiance, C/W 4 3.5 3 2.5 2 1.5 1 0 5 10 15 20 25 30 Tip ",eorcnce, mm Asympt-o-t-ic---fr-e--e space Sponwise spocinp= 5 mm -*.. - Spanwise spacing = 10 mm -.G..- Spanwise spacing = ? 5 mrr. -a-.- Spcnwise spacing = 25 mn Spanwise spacing = 35 mm - - Pi'igurc 7. Flow bypass effect on PowerPC 604 lieat sink resistances. 1 5.5 - Tip Cleoronce, mm - Asymptotic free space - _ _- --- - Spanwise spacing= 5 mm d- Sponwise spacing = 10 m m l(r- --- Spanwise spocing = 15 mm Ti; Spanwise spacing = 25 mm -_ Spanwise spacing = 35 mm Figure 6. Flow bypass effcct on PowerPC 604 heat sink prcssure drops. 3.5 I f 0 5 10 15 20 25 X, 5 Tip clearance, mm - Asymptotic free space spanwisespacing = 25 mm --U-- Spanwise spacing= 5 m m -.--*.-. - Spanwise spacing = 15 mm --e spanwisespacing = 40 mm -A-- Figure 8. NOWbypass effect on PowerPC 603 module thermal performance. 253 Thirteenth IEEE SEMI-THERMTMSymposium Uodule thcrtnal rosistance, C/’w h?odule thermc! resistance, C/w 4J 5 ! 4.5 I 4 t 3- 3.5 *-- A - 3 2’5 / 2.5 2 f 1 0 5 10 15 20 25 Thermal conductivity, \V/m k Card I( PowerPC 603 Mc!,~ Y Ccrd K powerpc 604 M C ! ~ -c Greose K PowerPC 6 0 3 MC!J ----!+--- Grease K PowerPC 604 L4Cb.4 ----e--. Figure 11. MCM card and grease themial co11ductivity cffccts on module thermal resistance. CIodule thermal resistcnce. C/w Module thermal resistance, c/Mf 0 100 200 300 Thermal conductivity, W/m k suSsDote K PuserDC 603 :?.CM ----ft scbstate K PoKer?C 6 0 4 MC!A Cap K PowerPC 603 GCM --~ _. . - . Cap K PowerPC 604 MCh! 1 400 2 1 1 , 1 0 1 2 3 4 5 6 Figure 10. MCM substrate and cap thermal Figure 12. MCM cap thickness effcct on module conductivity effects 011modulc thertnal resistance. thcrmal resistance. 254 Thirteenth IEEE SEMI-THERMm Symposium Perccntogc of hank rub, Z 100- 1 3ercen:cge o f heat rate, % I % 95 90 '60 I I I 0 50 100 150 Substrate thermal conductivity, Wm'! I 200 k 603 chip heat Into ccp - PowerPC 603 A!CC -a 604 chip heat raft into cap PowcrPC 604 UCM Module heat rate into heatsink Power.P-C-6-0*3-M-C-U Module heo: ra?e into heakink PowerPC 604 YCt.! I 6..0 0 5 10 15 20 25 Card thermal conductivity, W/" k - 603 chlp heat rate into cc? PowerPC 603 MCU 604 chip heat rate into c o p Power-PC.6-04 L{CM Module heat rate into heckink PowerPC 603 tic!!. ----e--- k!odule heot rate into hectsink PowerDC 604 K!.! __-_e--- 1;igurc 13. MChI Iicnt Iatc perccntage distribution duc to suhstratc hcr-mal conductivity. Figure 15. MCM heat ratc percentage distribution due to card thermal conductivity. Percentcge of heat rcte, Z 90 A===- 80 70 60 30 , I 0 50 100 150 200 Cop inermol conductivlty, \Y/" ?j - 603 chip heat rate into cap PowerPC 603 MCM - 604 chip heat rate into cap PowerPC 604 MCU Module heat r r t e into heatsink PowerPC 603 MCM ---*e--- Module h e o t rote into heatsink ---- -. -* PowerPC 604 UCU 50 0 0.5 1 1.5 2 2.5 3 Grease ihermal conductivity, \?'/" k 603 ,-hip heat rate into ccp --. PowerPC 603 MCM 604 chip heat rate into PowerPC 604 fAc!,i hiodule heat rate into h e o m k Po8erPC 603 UCU .--am..- E.!odu!e heat rcte into heatsink PowerPC 6 0 4 L!Ch! Figure 14. MCM heat ratc pcrccniagc distribution due to cap thcrmal conductivity. Figure 16. MCM heat rate pcrcentage distribution due to grease thermal conductivity. 255 Thirteenth IEEE SEMI-THERM" Symposium Percentago of hoot ratc, Z 1 05 - 80 . 65 I I , I 0 : 2 3 4 5 6 7 Ccp thickness, mm 503 rate inka ccp PowerPC GO3 hqC?! 60,1 chip rate intocap PowerPC 604 h!Ch! Yodule heat rate into heatsink PowerPC 603 MCM Module heat rate into heatsink _ _ _ _ PowerPC 604 K h ! c--- 256 Thirteenth IEEE SEMI-THERMTMSymposium

Top_arrow
回到顶部
EEWORLD下载中心所有资源均来自网友分享,如有侵权,请发送举报邮件到客服邮箱bbs_service@eeworld.com.cn 或通过站内短信息或QQ:273568022联系管理员 高进,我们会尽快处理。